Here is the error:
Error response from daemon: invalid mount config for type “bind”: stat /mnt/data: transport endpoint is not connected
Error: failed to start containers: storagenode
The storage on my node is all wrapped up into one directory and mounted to /mnt/cloud and it keeps getting disconnected.
I added more RAM and that stopped the disconnections for a while, however it is back.
Having the storage directory be remote is not a supported configuration and is likely to fail. You can find some hints on the forum what does it take to make it work, but you’re on your own here.
Sorry that you were misled by my naming of the directory. No, it’s not a remote server. I just named it cloud when I made the node. I changed the name to data, so people don’t get confused.
It’s not (just) the directory name. This error message (transport endpoint is not connected) basically never happens for any typical local file systems, like ext4, btrfs, etc., and it’s most commonly encountered when using fuse or a networked file system. What file system do you use on /mnt/data?
So, all of the drives individually are ext4. However, when bundled together it is fuse.mhddfs.
Then it seems mhddfs is not reliable enough, sorry. The usual recommendation is to run separate nodes on drives, can you do that?
No, I can’t unfortunately. Restarting the server fixes this problem and unmounting and remounting the mount point seems like it could also fix this problem. This is all based on what I have read online if anyone else reads this post. You could also downgrade mhddfs to 0.1.38 to fix this issue. And there is a github repo called mhddfs-nosegfault that fixes this issue.
If I replaced the original mhddfs with the custom package (mhddfs-nosegfault) in the /bin directory will the OS use the custom package?